An edible mushroom growth management method and system based on a knowledge graph

Abstract

The present application belongs to the technical field of edible mushroom growth management, and in particular relates to an edible mushroom growth management method and system based on a knowledge graph, which generates a cap abnormality recognition result by recognizing cap abnormality of a target edible mushroom in a multi-region edible mushroom mixed cultivation environment; obtains a potential correlation path corresponding to the abnormal result based on a constructed knowledge graph and a correlation path reasoning model, in combination with the cap abnormality recognition result; calculates a correlation strength score of each potential correlation path, and selects the path with the highest score as an abnormal cause-effect path of the target edible mushroom; and generates and executes an abnormal regulation strategy based on the abnormal cause-effect path. Through the present application, fine and intelligent management of the edible mushroom growth environment can be achieved, yield and quality loss caused by failure to recognize cross-region influence can be effectively avoided, and production efficiency and management level of the edible mushroom are significantly improved.

Description

Technical Field

[0001] This application relates to the field of edible fungi growth management technology, and more specifically, to a knowledge graph-based method and system for edible fungi growth management. Background Technology

[0002] In modern edible mushroom cultivation facilities, intelligent management systems utilize a built-in knowledge system to precisely manage the growth environment of edible mushrooms. This knowledge system includes suitable growth parameters for different mushroom varieties, as well as images of their health and diseases. The system monitors environmental data and the growth of the mushroom bed in real time using sensors and cameras, comparing this data with baseline information in the knowledge system to make corresponding adjustments. For example, when the humidity falls below the standard value, the system automatically activates the spray equipment; when yellow spots appear on the mushroom caps and match bacterial infection characteristics in the knowledge system, the system issues an early warning. In standardized mushroom houses for a single variety, this system operates well, effectively ensuring both yield and quality.

[0003] However, existing intelligent management systems face challenges in multi-regional mixed cultivation of edible fungi. To improve space utilization, large mushroom houses are often divided into multiple zones for cultivating different varieties of edible fungi. Although physical partitions separate the zones, the air circulation and temperature and humidity control systems throughout the entire mushroom house are largely shared, resulting in a lack of complete environmental isolation between different zones. Existing management systems typically treat the growth models of different varieties as independent units, managing each zone independently.

[0004] This independent management model may lead to cross-regional impacts. For example, when the spraying system is activated for shiitake mushrooms in zone A, which requires higher humidity, some of the humid air may drift into zone B, causing the local humidity in zone B's oyster mushroom growing environment to be slightly higher than its optimal level. At this time, the sensors in zone B may only detect minor fluctuations, but a faint, atypical grayish-white color may appear on the oyster mushroom caps. This color change is not drastic and does not match any known disease or contamination characteristics recorded in the knowledge system.

[0005] While the image processing component of the existing system can capture this color deviation, its knowledge base lacks the causal relationship between "slightly higher humidity" and the "pale grayish-white caps" of the oyster mushrooms. Therefore, the system cannot pinpoint the exact cause and categorize it as an "unknown variation" or "individual difference within the normal range," thus failing to issue an alert or suggest adjustments. In this situation, the system continues to manage each area independently as planned, failing to identify the negative impact of normal operations in area A on area B. Ultimately, the yield of oyster mushrooms in area B may be lower than expected, and the quality may decline. Although the existing system collects all relevant data, the limitations of its knowledge base prevent it from reasoning about and judging new problems arising from the interaction of different production processes within the same physical space. Summary of the Invention

[0006] This application discloses a knowledge graph-based method and system for managing the growth of edible fungi, which aims to solve the problem that existing intelligent management systems cannot effectively identify and infer the abnormal growth of edible fungi caused by cross-regional influences in multi-regional mixed cultivation scenarios, as well as the resulting decline in yield and quality.

[0007] This invention is achieved using the following technical solution: a knowledge graph-based method for managing the growth of edible fungi, comprising: In a multi-regional mixed cultivation environment for edible fungi, abnormalities in the caps of target edible fungi are identified, and corresponding cap abnormality identification results are generated. Based on knowledge graphs, a correlation path reasoning model for the corresponding target edible fungi is constructed. Based on the cap anomaly identification results and the correlation path reasoning model, the potential correlation paths corresponding to the cap anomaly identification results are obtained. Calculate the association strength score for each potential association path, and select the potential association path with the highest association strength score as the abnormal causal path of the target edible fungus. Based on abnormal causal paths, an abnormal control strategy is generated and adjusted according to the abnormal control strategy.

[0008] Through this technical solution, this application can effectively identify abnormal conditions of edible fungi caps in multi-regional mixed cultivation environments, and use knowledge graphs to infer the potential causal paths leading to the abnormalities, thereby generating precise control strategies. This solves the problem in existing technologies that cannot identify cross-regional influences causing abnormalities in edible fungi, and improves the intelligence and precision of edible fungi management.

[0009] Preferably, in a multi-regional mixed cultivation environment for edible fungi, abnormalities in the caps of the target edible fungi are identified, and corresponding cap abnormality identification results are generated, including: Acquire panoramic images of the target mushroom bed for the target edible fungi; The panoramic image of the target mushroom bed is input into a pre-trained target edible mushroom cap anomaly recognition model. The output results are used to determine whether there are any areas with abnormal cap color in the panoramic image of the target mushroom bed. The pre-trained target edible fungus cap anomaly recognition model is obtained by training the target edible fungus cap anomaly recognition model on images of healthy target edible fungus caps and images of target edible fungus caps with identified diseased and abnormal caps. For a panoramic image of a target mushroom bed with abnormal cap color, determine the number of abnormal cap color areas, the color deviation value of each abnormal cap color area, the proportion of abnormal areas, the duration, and the sub-image of each abnormal cap color area, and generate the cap anomaly recognition result.

[0010] Through this technical solution, this application can use image recognition technology to accurately identify and quantify abnormalities in the caps of edible fungi, providing accurate and comprehensive data support for subsequent causal path reasoning, and effectively improving the accuracy and comprehensiveness of anomaly identification.

[0011] Preferably, the association path reasoning model of the target edible fungus is a graph database query engine that includes the knowledge system of the target edible fungus; wherein, the nodes in the knowledge system of the target edible fungus are the variety, growth environment parameters, regulatory behavior, type and location of regulatory equipment, and biological characteristics of the target edible fungus; the edges in the knowledge system of the target edible fungus are the association relationships between different nodes, and the edges formed by connecting the same node to other different nodes have different weights.

[0012] Through this technical solution, this application constructs a structured and reasonable knowledge graph, which links multi-dimensional information related to the growth of edible fungi, providing a solid foundation for subsequent causal path reasoning and enabling the system to perform deeper logical analysis.

[0013] Preferably, based on the cap anomaly identification results and the association path reasoning model, the potential association paths corresponding to the cap anomaly identification results are obtained, including: In the association path reasoning model, the node that is related to the result of cap anomaly identification is selected as the starting node; Perform a reverse lookup within the knowledge system to find growth environment parameters that are associated with the starting node, which serve as the first potential intermediate node; For each first potential intermediate node obtained by reverse query, the reverse query in the knowledge system causes or affects the regulatory behavior of the corresponding first potential intermediate node, which is then used as a second potential intermediate node. Based on each second potential intermediate node obtained from the query, a reverse query is performed in the knowledge system to determine the control device and its location that executes the second potential intermediate node, which serves as the termination node. Potential associated paths include a termination node, a second potential intermediate node, a first potential intermediate node, and a starting node.

[0014] Through this technical solution, this application can use reverse query to trace back from abnormal cap phenomena to possible environmental parameters, regulatory behaviors, and regulatory devices, thereby systematically constructing all possible causal chains and providing comprehensive candidate paths for subsequent abnormality location.

[0015] Preferably, in the step of calculating the association strength score for each potential association path, the calculation logic includes: If the number of nodes in a potential association path is less than the preset first number of nodes, then the first association strength score for the corresponding potential association path is the first preset association strength score. If the number of nodes in a potential association path is greater than or equal to the preset first number of nodes and less than the preset second number of nodes, then the first association strength score of the corresponding potential association path is given as the second preset association strength score. If the number of nodes in a potential association path is greater than the preset second number of nodes, then the first association strength score of the corresponding potential association path is given as the third preset association strength score; wherein, the first preset association strength score is greater than the second preset association strength score, which is greater than the third preset association strength score. In the knowledge system, a preset node confidence level is set for each node, and a preset edge confidence level is set for each edge. The second association strength score of the potential association path is the product of the preset node confidence level of all nodes and the preset edge confidence level of each edge. Compare the occurrence time of cap abnormality with the occurrence time of the regulatory behavior of the second potential intermediate node. If the occurrence time of the regulatory behavior of the second potential intermediate node is within a first preset time range before the occurrence time of cap abnormality, the third association strength score of the potential association path is given as the time matching score calculated based on the occurrence time of cap abnormality and the occurrence time of the regulatory behavior of the second potential intermediate node. Within a second preset time range before and after the abnormality of the cap of the target edible fungus, whether the control device in the area outside the area where the target edible fungus is located exhibits control behavior. If it does, the fourth association strength score of the potential association path is given as the control association score. If it does not, the fourth association strength score of the potential association path is given as 0. The calculated association strength score for any potential association path is equal to the first association strength score plus the second association strength score plus the third association strength score plus the fourth association strength score.

[0016] Through this technical solution, this application comprehensively considers multiple dimensions such as path length, confidence of nodes and edges, time matching degree, and cross-regional regulation impact to quantitatively evaluate potential related paths, enabling the system to more accurately identify the most likely abnormal causal paths and improve the reliability of causal reasoning.

[0017] Preferably, the time difference obtained by subtracting the time point of the abnormal occurrence of the cap from the time point of the occurrence of the regulatory behavior of the second potential intermediate node is multiplied by a preset attenuation coefficient, and then an exponential operation with the natural constant as the base is performed to obtain the time matching degree score; if the time difference between the time point of the abnormal occurrence of the cap and the time point of the occurrence of the regulatory behavior of the second potential intermediate node is greater than 72 hours, the corresponding potential association path is abandoned.

[0018] Through this technical solution, this application introduces a time decay mechanism, which makes the influence of time-closer regulatory behaviors on cap abnormalities more weighted, and sets a time threshold to effectively exclude time-irrelevant regulatory behaviors, further improving the accuracy of causal inference.

[0019] Preferably, an anomaly control strategy is generated based on the abnormal causal path, and adjustments are made according to the anomaly control strategy, including: Based on each node in the abnormal causal path, identify the node that causes the abnormal cap color; An abnormal control strategy is generated based on the node content that causes the abnormal cap color on the abnormal causal path. The abnormal control strategy is sent to the control equipment in the corresponding area, and the control equipment is controlled to perform control according to the abnormal control strategy.

[0020] Through this technical solution, this application can directly locate the specific node causing the abnormality based on the identified abnormal causal path, and generate targeted control strategies to achieve precise intervention in the growth environment of edible fungi, effectively solving the abnormality problem.

[0021] Preferably, after the step of adjusting according to the abnormal control strategy, the method further includes a step of monitoring the growth of the target edible fungus, including: A panoramic image of the target edible fungus's mushroom bed is obtained. The panoramic image of the target edible fungus's mushroom bed is then input into a pre-trained target edible fungus cap anomaly recognition model. The output results are used to determine whether there are any abnormal cap color areas in the panoramic image of the target edible fungus's mushroom bed. The pre-trained target edible fungus cap anomaly recognition model is obtained by training the target edible fungus cap anomaly recognition model on images of healthy caps of the target edible fungus and images of caps of the target edible fungus that have been identified as diseased and abnormal. If, after the third preset time following the implementation of the control strategy, the abnormal color area of ​​the cap in the panoramic image of the target edible fungus disappears, then the monitoring of the growth of the target edible fungus will be stopped. If the abnormal color area of ​​the cap in the panoramic image of the target edible fungus bed does not disappear after the third preset time following the implementation of the control strategy, an alarm signal will be issued.

[0022] This technical solution introduces a feedback monitoring mechanism for the control effect, which can promptly assess the effectiveness of the control strategy and decide whether to continue monitoring or issue an alarm based on the monitoring results, thus forming a closed-loop management system and ensuring the complete resolution of abnormal problems.

[0023] Preferably, after the step of adjusting according to the abnormal control strategy, the method further includes: The abnormal causal path of the target edible fungus is added to the association path reasoning model in the form of nodes and edges; among them, The path formed by connecting the starting node and the first potential intermediate node in the abnormal causal path of the target edible fungus is added to the association path inference model as a first-level association path. The path formed by connecting the first potential intermediate node and the second potential intermediate node in the abnormal causal path of the target edible fungus is added to the association path inference model as a secondary association path.

[0024] Through this technical solution, this application realizes the dynamic updating and learning of the knowledge graph, adding successfully identified abnormal causal paths to the knowledge system, enabling the system to continuously accumulate experience, improve the accuracy and efficiency of future anomaly identification and reasoning, and achieve self-growth of knowledge.

[0025] This invention also discloses a knowledge graph-based edible fungus growth management system, comprising: The anomaly identification module is used to identify abnormalities in the caps of target edible fungi in a multi-regional mixed cultivation environment and generate corresponding cap anomaly identification results. The potential association path determination module is used to construct an association path reasoning model for the corresponding target edible fungus based on the knowledge graph, and obtain the potential association path corresponding to the cap anomaly identification result based on the cap anomaly identification result and the association path reasoning model. The abnormal causal path determination module is used to calculate the association strength score of each potential association path and select the potential association path with the highest association strength score as the abnormal causal path of the target edible fungus. The regulation module is used to generate anomaly regulation strategies based on abnormal causal paths and to make adjustments according to the anomaly regulation strategies.

[0026] This application provides a system for implementing the above-mentioned method through this technical solution. Through modular design, each functional unit is clearly divided, making the system structure clear, easy to implement and maintain, and able to effectively execute the knowledge graph-based edible fungus growth management method.

[0027] Compared with the prior art, the beneficial effects of the present invention are: This application discloses a knowledge graph-based method for managing the growth of edible fungi. In a multi-regional mixed cultivation environment for edible fungi, it first identifies abnormalities in the caps of target edible fungi, generating cap anomaly identification results. Then, based on the constructed knowledge graph and association path reasoning model, combined with the cap anomaly identification results, potential association paths corresponding to the anomaly results are obtained. Next, the association strength score of each potential association path is calculated, and the path with the highest score is selected as the causal path of the anomaly for the target edible fungi. Finally, based on this causal path, an anomaly control strategy is generated and executed.

[0028] This method effectively addresses the problem in existing technologies for multi-regional mixed cultivation of edible fungi. Due to the non-isolated environments of different regions, cross-regional influences between different fungi varieties can occur, and the limitations of existing management systems prevent them from reasoning about and judging new problems arising from these interactions. This leads to an inability to identify the true cause of atypical cap anomalies, resulting in decreased yield and quality. This application, by introducing knowledge graphs and association path reasoning mechanisms, can reverse-engineer the underlying environmental parameters, regulatory behaviors, and equipment causing the anomalies from the cap anomalies themselves, and even identify hidden cross-regional influences, thereby accurately locating the cause of the anomalies. Quantitative evaluation of the correlation strength of potential association paths ensures the accuracy of the identified causal paths. Finally, based on the precise causal paths, targeted regulatory strategies are generated, achieving refined and intelligent management of the edible fungi growth environment. This effectively avoids yield and quality losses caused by the failure to identify cross-regional influences, significantly improving the production efficiency and management level of edible fungi. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 This is a flowchart illustrating a knowledge graph-based method for managing the growth of edible fungi, as provided by this invention.

[0031] Figure 2 This is a schematic diagram of the structure of an edible fungus growth management system based on knowledge graphs provided by the present invention. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] In multi-regional mixed cultivation environments for edible fungi, traditional intelligent management systems face significant challenges in handling cross-regional impacts. Existing systems typically manage the growth models of different varieties as independent units, resulting in their knowledge base being unable to effectively reason about and assess new problems arising from the interaction of different production processes within the same physical space. For example, when normal operations in one region have a minor but persistent negative impact on another region, existing systems often fail to identify the exact causal relationship, thus failing to take timely and effective control measures, potentially leading to a decline in the yield and quality of edible fungi.

[0034] like Figure 1 As shown, this application proposes a knowledge graph-based method for the management of edible fungi growth, including: S110: In a multi-regional mixed cultivation environment for edible fungi, identify abnormalities in the caps of target edible fungi and generate corresponding cap abnormality identification results.

[0035] S120: Based on the knowledge graph, construct the association path reasoning model for the corresponding target edible fungus. Based on the cap anomaly identification results and the association path reasoning model, obtain the potential association path corresponding to the cap anomaly identification results.

[0036] S130: Calculate the association strength score for each potential association path, and select the potential association path with the highest association strength score as the abnormal causal path of the target edible fungus.

[0037] S140: Based on the abnormal causal path, generate an abnormal control strategy and adjust it according to the abnormal control strategy.

[0038] This application, by constructing a knowledge graph and an association path reasoning model, can deeply explore the complex causal relationship between abnormal caps of edible fungi and potential environmental factors and regulatory behaviors in multi-regional mixed cultivation environments, thereby achieving accurate identification and effective control of abnormal situations and significantly improving the growth management level of edible fungi.

[0039] To better understand the technical solution proposed in this application, some key terms and implementation environments involved will be explained first.

[0040] The multi-regional mixed cultivation environment for edible fungi referred to in this application is a cultivation scenario in which a large mushroom house is divided into multiple areas and different varieties of edible fungi are grown in each area. Although there is physical isolation, the air circulation, temperature and humidity control systems of each area are shared to some extent, which may lead to cross-regional environmental impacts.

[0041] Targeted edible fungi are varieties of edible fungi that require key monitoring and management during a specific management cycle.

[0042] Abnormalities in mushroom caps refer to deviations from the normal growth state in terms of color, shape, and texture of edible mushroom caps, such as abnormal color, spots, and deformities.

[0043] Knowledge graphs are a structured form of knowledge representation that describes entities and the relationships between them through nodes and edges, and can effectively store and reason about complex relational information.

[0044] The association path reasoning model is a computational model based on knowledge graphs used to infer the causal chain between cap abnormalities and potential causes.

[0045] Specifically, there are several methods for identifying abnormalities in the caps of target edible fungi. For example, one method involves manual inspection, where experienced agronomists periodically examine the caps of the target fungi. If any abnormalities are found in color, shape, or other aspects, they are recorded and an anomaly identification result is generated. This method relies on human experience and offers some flexibility. Another method involves deploying high-resolution cameras to periodically or in real-time capture images of the target edible fungi's caps. These images are then manually analyzed to identify anomalies and generate identification results. This method reduces the frequency of on-site inspections but still requires significant manual interpretation.

[0046] There are various methods for constructing association path reasoning models. For example, a large amount of edible fungus growth data, environmental parameters, control records, and expert experience can be manually collected. Then, professionals can organize this information into structured knowledge and manually construct nodes and edges in a knowledge graph, forming a graph database containing information such as the target edible fungus variety, growth environment parameters, control behaviors, control equipment type and location, and biological characteristics. Another approach is to utilize natural language processing and machine learning techniques to automatically extract entities and relationships from large amounts of unstructured text data (such as research papers, industry reports, and expert interview records) and map them to a predefined ontology structure, thus semi-automatically constructing a knowledge graph.

[0047] After obtaining the potential association paths, this application needs to calculate the association strength score for each potential association path and select the potential association path with the highest association strength score as the abnormal causal path of the target edible fungus. There are several ways to calculate the association strength score. For example, a score can be simply assigned based on the number of nodes in the path; the shorter the path and the fewer the nodes, the higher the association strength is considered. Alternatively, different weights can be assigned based on the type of edges in the path; for example, edges representing a "cause" relationship have a higher weight than edges representing a "influence" relationship, and then the weights of all edges in the path are summed to obtain the association strength score.

[0048] There are several ways to generate anomaly control strategies. For example, a rule base can be pre-set. When a specific abnormal causal path is identified, the system generates a corresponding control strategy based on the pre-set control schemes in the rule base. For instance, if an abnormal causal path indicates that excessive humidity is causing abnormal cap color, a control strategy to reduce humidity will be generated. Another approach is for the system to intelligently recommend or generate the optimal control strategy based on the node information involved in the abnormal causal path, querying historical control data and expert experience databases. For instance, if it is found that the increased humidity in area B is caused by spraying behavior in area A, the system may suggest adjusting the spraying time and intensity in area A, or increasing local ventilation in area B.

[0049] The overall working principle of this application lies in connecting seemingly independent events in a multi-regional mixed cultivation environment of edible fungi by constructing and utilizing a knowledge graph, thereby revealing deep-seated causal relationships. When the caps of target edible fungi exhibit abnormalities, traditional systems may only be able to identify the abnormal phenomena but struggle to trace their root causes. This application first obtains specific abnormal information by identifying the abnormal cap conditions. Subsequently, using a pre-constructed knowledge graph and a relational path reasoning model, the system can infer all possible potential relational paths from the abnormal cap phenomena. These paths connect the cap abnormalities with various environmental parameters, regulatory behaviors, and equipment. For example, when the caps of oyster mushrooms in area B exhibit a pale grayish-white abnormality, the system can not only identify this abnormality but also query the knowledge graph for growth environment parameters related to "pale grayish-white oyster mushroom caps," such as "high humidity." Further queries can then be made for regulatory behaviors that lead to "high humidity," such as "spraying in area A," ultimately locating the equipment and location executing this regulatory behavior.

[0050] After obtaining all potential association paths, the system calculates the association strength score for each path. This score comprehensively considers multiple factors such as path length, node and edge confidence, time matching degree, and cross-regional regulatory association, thereby quantifying the causal strength between each path and the cap abnormality. By selecting the path with the highest association strength score, the system can accurately locate the most likely root cause of the cap abnormality, i.e., the abnormal causal path. For example, if the path "spraying in area A leads to high humidity in area B, which in turn causes the oyster mushroom caps in area B to be pale grayish-white" has the highest association strength score, then the system can determine that this is the root cause of the oyster mushroom abnormality in area B.

[0051] Once the causal path of the anomaly is identified, the system can generate targeted anomaly control strategies based on that path. For example, in the case above, the system might generate a control strategy of "adjusting the spraying time or intensity in area A and increasing local ventilation in area B." Finally, the system sends these strategies to the corresponding control devices and controls the devices to adjust according to the strategies. The entire process forms a closed loop, from anomaly identification to cause localization, and then to strategy generation and execution, achieving intelligent and precise management of abnormal edible fungi growth in multi-region mixed cultivation environments.

[0052] This application offers significant advantages and innovations compared to existing technologies. Traditional systems, in multi-regional mixed cultivation environments, struggle to identify and handle complex cross-regional causal relationships due to the limitations of their knowledge base. They often attribute atypical cap abnormalities to "unknown variations" or "individual differences within the normal range," thus missing optimal control opportunities. The core innovation of this application lies in the introduction of knowledge graphs and association path reasoning models. This allows the system to overcome the limitations of traditional knowledge systems and make explicit the potential relationships between different regions, varieties, and control behaviors.

[0053] Specifically, this application constructs a graph database query engine containing a knowledge system of the target edible fungi. Information such as variety, growth environment parameters, regulatory behaviors, regulatory equipment, and biological characteristics are linked in the form of nodes and edges, and assigned different weights. This allows for a more comprehensive and detailed description of the complexity of the edible fungi's growth environment. When abnormalities occur in the cap, the system no longer relies solely on pre-defined single disease feature matching, but can explore all possible causal chains within the knowledge system through reverse queries, generating potential correlation paths. This graph-based reasoning capability enables the system to discover implicit correlations that are difficult to capture with traditional rule bases or single models, such as the small but persistent impact of normal operation in area A on area B.

[0054] Furthermore, this application quantifies potential correlation paths by calculating correlation strength scores, comprehensively considering multiple dimensions such as path length, node confidence, edge confidence, time matching degree, and cross-regional regulatory correlation. This multi-dimensional evaluation mechanism enables the system to accurately screen the most likely abnormal causal paths from numerous potential causes, avoiding misjudgments and ineffective regulation. For example, even if multiple factors may lead to cap abnormalities, the system can identify the most direct and strongest causal relationship through correlation strength scores.

[0055] Ultimately, based on the accurately identified causal paths of anomalies, this application can generate highly targeted anomaly control strategies and guide the adjustment of control equipment. This complete closed loop, from "identifying anomalies" to "locating the cause" and then to "precise control," significantly improves the intelligence and precision of edible fungi growth management, effectively solving the problem of cross-regional impact identification and handling faced by traditional systems in multi-regional mixed cultivation scenarios, thereby ensuring the yield and quality of edible fungi.

[0056] In other embodiments of the present invention, in a multi-regional mixed cultivation environment for edible fungi, abnormalities in the caps of target edible fungi are identified, and corresponding cap abnormality identification results are generated, including: Obtain a panoramic image of the target mushroom bed for the target edible fungus; The panoramic image of the target mushroom bed is input into a pre-trained target edible mushroom cap anomaly recognition model. The output result is used to determine whether there are abnormal cap color areas in the panoramic image of the target mushroom bed. The pre-trained target edible mushroom cap anomaly recognition model is obtained by training the target edible mushroom cap anomaly recognition model on images of healthy target edible mushroom caps and images of target edible mushroom caps with identified diseased and abnormal caps. For a panoramic image of a target mushroom bed with abnormal cap color, determine the number of abnormal cap color areas, the color deviation value of each abnormal cap color area, the proportion of abnormal areas, the duration, and the sub-image of each abnormal cap color area, and generate the cap anomaly recognition result.

[0057] Specifically, acquiring panoramic images of the target edible fungus's cultivation bed involves periodically acquiring complete image data of the bed, either periodically or under specific triggering conditions, using image acquisition devices deployed in the edible fungus cultivation environment, such as high-definition cameras or vision systems mounted on drones. These images should clearly reflect the visual characteristics of the cap, including its color and shape. The panoramic image of the target cultivation bed is then input into a pre-trained target edible fungus cap anomaly recognition model. The model's output determines whether there are any abnormal cap color regions within the panoramic image. This can be understood as a deep learning-based image recognition model, such as a convolutional neural network (CNN). This model is trained to identify the normal color range of the cap and various abnormal color patterns. When the panoramic image of the target cultivation bed is input, the model analyzes the pixel information in the image and outputs whether there are regions deviating from the normal color range, thus determining whether there are cap color anomalies. In practical applications, the pre-trained target edible mushroom cap anomaly recognition model is trained using images of healthy target edible mushroom caps and images of identified diseased and abnormal target edible mushroom caps. Specifically, the model's training dataset includes a large number of labeled images of healthy target edible mushroom caps and images of identified diseased and abnormal target edible mushroom caps. Through supervised learning, the model learns the visual features that distinguish healthy and abnormal caps, aiming to improve its recognition accuracy. Further, for panoramic images of target mushroom beds containing cap color anomalies, the system determines the number of cap color anomalies, the color deviation value of each anomaly, the proportion of anomalies, the duration of anomalies, and the sub-images of each anomaly, generating cap anomaly recognition results. Once the model identifies cap color anomaly regions, the system performs quantitative analysis on these regions. For example, image segmentation techniques are used to determine the boundaries of abnormal regions and calculate their number; color deviation values ​​are calculated by comparing them with the average color of healthy caps; the proportion of abnormal regions is calculated by the ratio of the pixel area of ​​the abnormal region to the total cap area; the duration of the abnormality is recorded through continuous monitoring; and sub-images of each abnormal region are captured for subsequent detailed analysis or manual verification. All this quantitative information collectively constitutes the cap anomaly identification results, providing detailed data support for subsequent causal path reasoning.

[0058] The proposed solution first acquires panoramic images of the target mushroom bed, providing raw data for subsequent image analysis. These images are then input into a well-trained cap anomaly recognition model. This model, based on features learned from a large number of healthy and abnormal cap images, automatically and efficiently determines the presence of regions with abnormal cap color. Once an anomaly is detected, the system further refines and quantifies the anomaly region, including determining its quantity, color deviation value, proportion, duration, and extracting sub-images of the anomaly region. This detailed quantitative analysis ensures the comprehensiveness and accuracy of the cap anomaly recognition results, providing rich and reliable input data for subsequent anomaly causal path reasoning.

[0059] In some embodiments of this application, in order to more specifically realize the construction of a correlation path reasoning model for the target edible fungus based on a knowledge graph, and to obtain the potential correlation path corresponding to the cap anomaly identification result based on the model, the above-mentioned correlation path reasoning model has been further defined.

[0060] The association path reasoning model for the target edible fungi is set up as a graph database query engine that includes the target edible fungi's knowledge system. Nodes in the target edible fungi's knowledge system are defined as the variety, growth environment parameters, regulatory behaviors, type and location of regulatory equipment, and biological characteristics of the target edible fungi. Edges in the target edible fungi's knowledge system are defined as the association relationships between different nodes, and edges connecting the same node to other different nodes have different weights.

[0061] Furthermore, a graph database query engine is a database system specifically designed for storing, managing, and querying graph-structured data. Its core strength lies in its ability to efficiently handle complex relationships between nodes and edges, thereby supporting complex path queries and pattern matching. By constructing the knowledge system of the target edible fungi as a graph database, various entities (nodes) and their interactions (edges) can be intuitively represented.

[0062] In this knowledge graph, nodes are the basic units, representing various entities or concepts in the field of edible mushroom growth management. For example, variety nodes may include "shiitake mushroom" or "oyster mushroom"; growth environment parameter nodes may include "temperature," "humidity," or "carbon dioxide concentration"; control behavior nodes may include "ventilation," "humidification," or "cooling"; control equipment type and location nodes may include "fan in zone 1" or "humidifier in zone 2"; and biological characteristic nodes may include "mycelial growth rate" or "cap color." These nodes together constitute a comprehensive information view of the target edible mushroom's growth management.

[0063] Edges represent the relationships between these nodes. For example, there might be an edge between the variety node "shiitake mushroom" and the growth environment parameter node "temperature" with the concept of "suitable growth temperature"; similarly, there might be an edge between the regulatory behavior node "ventilation" and the growth environment parameter node "carbon dioxide concentration" with the concept of "reduction". To reflect the importance or strength of different relationships, edges connecting the same node to other nodes are assigned different weights. For instance, a certain regulatory behavior might have a greater impact on one environmental parameter than on another, which can be reflected by the edge weights. These weights can be set during the knowledge graph construction based on domain expert experience or historical data analysis.

[0064] This application's solution concretizes the association path reasoning model into a graph database query engine containing a knowledge system of the target edible fungi, effectively organizing and managing massive amounts of heterogeneous data in the field of edible fungi growth management. Upon receiving anomaly identification results for the cap, the graph database query engine leverages its powerful graph traversal and pattern matching capabilities to efficiently reverse-search and infer the potential association paths leading to the anomaly within a pre-built knowledge system, starting from nodes related to the cap anomaly. Through explicit definitions of nodes and edges, and the introduction of edge weights, the path reasoning process becomes more accurate and intelligent, enabling the selection of the most probable causal chains from complex association networks.

[0065] Based on the edible fungi growth management method of the present invention, potential association paths corresponding to the results of cap anomaly identification are obtained, including: In the association path reasoning model, the node that is related to the result of cap anomaly identification is selected as the starting node; Perform a reverse lookup within the knowledge system to find growth environment parameters that are associated with the starting node, which serve as the first potential intermediate node; For each first potential intermediate node obtained by reverse query, the reverse query in the knowledge system causes or affects the regulatory behavior of the corresponding first potential intermediate node, which is then used as a second potential intermediate node. Based on each second potential intermediate node obtained from the query, a reverse query is performed in the knowledge system to determine the control device and its location that executes the second potential intermediate node, which serves as the termination node. Potential associated paths include a termination node, a second potential intermediate node, a first potential intermediate node, and a starting node.

[0066] In this approach, nodes related to the anomaly identification results in the association path reasoning model are selected as starting nodes. This involves mapping the anomaly information contained in the anomaly identification results, such as the number of abnormal color areas on the cap, color deviation values, the proportion of abnormal areas, and duration, to corresponding nodes in the knowledge system. These nodes, as the starting points of potential causal chains, represent the observed anomalies.

[0067] Reverse lookup traces the direct cause (growth environment parameters) from the result (abnormal cap). Growth environment parameters can include, but are not limited to, temperature, humidity, carbon dioxide concentration, light intensity, and ventilation. These parameters directly affect the growth status of edible fungi and the health of the caps. By querying, all growth environment parameters that may have caused or affected the starting node can be identified and used as the first potential intermediate nodes.

[0068] Based on this, for each first potential intermediate node obtained through reverse lookup, the regulatory behaviors that cause or affect the corresponding first potential intermediate node are identified as second potential intermediate nodes through reverse lookup within the knowledge system. Regulatory behaviors refer to various operations taken to change or maintain growth environment parameters, such as turning humidifiers, fans, heaters, supplemental lighting, etc., or adjusting their operating parameters. This step identifies which regulatory behaviors might lead to abnormalities in the growth environment parameters represented by the first potential intermediate node.

[0069] Subsequently, based on each second potential intermediate node obtained from the query, the knowledge system is used to reverse-search the control equipment and its location that executes the second potential intermediate node, serving as the termination node. Control equipment refers to the physical device that performs the control behavior, such as humidifiers, ventilation equipment, heating equipment, and supplemental lighting equipment. The location specifies the exact installation position of the equipment in a multi-regional mixed cultivation environment for edible fungi. The termination node represents the final physical source of the potential causal chain, i.e., the specific control equipment and its operating area.

[0070] Thus, the potential correlation path was constructed as a complete chain from the terminating node, the second potential intermediate node, the first potential intermediate node back to the starting node. This path clearly depicts the entire causal process from the operation of specific control equipment, through control behaviors, affecting growth environment parameters, and ultimately leading to abnormal cap growth.

[0071] The proposed solution, through the detailed steps outlined above, enables systematic reverse reasoning using a knowledge graph, starting from observed abnormalities in the mushroom cap. By tracing back layer by layer, it delves deeper from the apparent anomaly (starting node) to its direct environmental impact (first potential intermediate node), then to the regulatory operations leading to these environmental impacts (second potential intermediate node), and finally locates the specific equipment and its location performing these operations (terminating node). This structured reverse query mechanism clearly presents the potential causal chain, providing a comprehensive set of candidate paths for subsequent identification of abnormal causal paths.

[0072] This application proposes a method for calculating the correlation strength score of potential correlation paths. However, in its implementation, if only a single or simple calculation logic is used, it may not be able to comprehensively and accurately assess the true causal strength of potential correlation paths. For example, if the complexity of the path, the confidence level of each element in the knowledge system, the temporal matching, and the potential cross-regional impact are not fully considered, it may lead to misjudgment of abnormal causal paths, thereby generating an inaccurate anomaly control strategy.

[0073] In this regard, this application further proposes that the calculation logic for the step of calculating the association strength score for each potential association path includes: If the number of nodes in a potential association path is less than the preset first number of nodes, then the first association strength score of the corresponding potential association path is given as the first preset association strength score; if the number of nodes in a potential association path is greater than or equal to the preset first number of nodes and less than the preset second number of nodes, then the first association strength score of the corresponding potential association path is given as the second preset association strength score. If the number of nodes in a potential association path is greater than the preset second number of nodes, then the first association strength score of the corresponding potential association path is given as the third preset association strength score; wherein, the first preset association strength score is greater than the second preset association strength score, which is greater than the third preset association strength score. In the knowledge system, a preset node confidence level is set for each node, a preset edge confidence level is set for each edge, and the second association strength score of the potential associated path is the product of the preset node confidence level of all nodes and the preset edge confidence level of each edge. Compare the occurrence time of cap abnormality with the occurrence time of the regulatory behavior of the second potential intermediate node. If the occurrence time of the regulatory behavior of the second potential intermediate node is within a first preset time range before the occurrence time of cap abnormality, the third association strength score of the potential association path is given as the time matching score calculated based on the occurrence time of cap abnormality and the occurrence time of the regulatory behavior of the second potential intermediate node. Within a second preset time range before and after the abnormality of the cap of the target edible fungus, whether the control device in the area outside the area where the target edible fungus is located exhibits control behavior. If it does, the fourth association strength score of the potential association path is given as the control association score. If it does not, the fourth association strength score of the potential association path is given as 0. The calculated association strength score for any potential association path is equal to the first association strength score plus the second association strength score plus the third association strength score plus the fourth association strength score.

[0074] The computational logic aims to comprehensively quantify the causal strength of potential association paths through multi-dimensional evaluation. The first association strength score is set based on the number of nodes in the potential association path. The number of nodes in a path can reflect its complexity or directness; for example, a shorter path may imply a more direct causal relationship, while a longer path may involve more intermediate steps. By setting a first, second, and third preset association strength score in descending order, different initial weights can be assigned to paths of different lengths, thus reflecting the impact of path length on causal strength.

[0075] The second association strength score considers the confidence level of nodes and edges in the knowledge system. In a knowledge graph, different nodes (such as species, environmental parameters, regulatory behaviors, etc.) and edges (associations) may have different reliability or importance. By setting preset node confidence and preset edge confidence for each node and each edge, and multiplying these confidence levels, the overall reliability of potential association paths can be quantified. Paths with higher confidence are more likely to be anomalous causal paths.

[0076] The third association strength score incorporates considerations of time matching. The occurrence of cap anomalies is often closely related to prior regulatory behaviors or environmental changes. By comparing the time of cap anomaly occurrence with the time of the regulatory behavior at the second potential intermediate node and calculating the time matching score, the temporal correlation between the regulatory behavior and the anomaly can be assessed. If the regulatory behavior occurs within a reasonable timeframe before the anomaly occurs, then that regulatory behavior is more likely to be causally related to the anomaly.

[0077] The fourth association strength score is used to assess the impact of cross-regional regulatory behavior. In multi-regional mixed cultivation environments, regulatory behavior in one region may sometimes affect the growth of edible fungi in adjacent or related regions. By verifying whether regulatory devices outside the region where the target edible fungus is located exhibit regulatory behavior and assigning a regulatory association score accordingly, potential cross-regional causal chains can be identified, avoiding the omission of important external influencing factors.

[0078] Ultimately, the association strength score of any potential association path is obtained by summing the scores from the four dimensions mentioned above. This comprehensive calculation method makes the evaluation of potential association paths more complete and accurate.

[0079] This application's solution constructs a multi-dimensional logic for calculating association strength scores by comprehensively considering the length of potential association paths, the confidence level of various elements in the knowledge system, temporal matching, and potential cross-regional impact. This method can more precisely quantify the causal strength of each potential association path, thereby avoiding misjudgments that may be caused by a single evaluation standard. By integrating information from different dimensions, this solution can more accurately identify the true causal path leading to abnormalities in edible mushroom caps, providing a solid foundation for the subsequent generation of anomaly control strategies.

[0080] In an embodiment of the present invention, it is assumed that in a certain edible fungus cultivation area, the cap of a target edible fungus exhibits abnormal coloration on a certain day. The system identifies three potential association paths using a knowledge graph: Path A: Environmental parameters (temperature too high) -> Control behavior (heating equipment failure) -> Control device (heater 100) Path B: Environmental parameters (humidity too low) -> Control behavior (humidifier not started) -> Control device (humidifier 200) Path C: Environmental parameters (insufficient light intensity) -> Adjustment behavior (damaged fill light) -> Adjustment equipment (fill light 300) Calculate the association strength score for path A: 1. First Association Strength Score (Based on Number of Nodes): Path A contains 3 nodes (environmental parameters, control behavior, control equipment). Assume the preset number of the first node is 2, and the preset number of the second node is 4. Since 3 is greater than or equal to 2 and less than 4, the first association strength score assigned to path A is the second preset association strength score (for example, set to 0.7).

[0081] 2. Second Association Strength Score (Based on Confidence): Assuming that in the knowledge system, the confidence score of the node with excessively high temperature is 0.9, the confidence score of the edge indicating heating equipment failure is 0.8, and the confidence score of the node indicating heater 100 is 0.95, then the second association strength score of path A is 0.9 * 0.8 * 0.95 = 0.684.

[0082] 3. Third Association Strength Score (Based on Time Matching): Assume the time of occurrence of the cap anomaly is T0, and the time of occurrence of the heating equipment malfunction control behavior (or event) is T0-24 hours. If the first preset time range is 48 hours, and the calculated time matching score is 0.9, then the third association strength score of path A is 0.9.

[0083] 4. Fourth Association Strength Score (Based on Regulation Association Score): Within a second preset time range before and after the occurrence of cap abnormality, verify whether the regulation equipment in the area outside the target edible fungus area exhibits regulation behavior. If it is found that a heater in an adjacent area also malfunctions during T0-30 hours, and this malfunction may affect the temperature of the target area through the ventilation system, then the fourth association strength score of path A is assigned as the regulation association score (e.g., set to 0.5).

[0084] Finally, the association strength score of path A is 0.7 + 0.684 + 0.9 + 0.5 = 2.784.

[0085] By performing similar calculations on all potential related paths, the path with the highest score is selected as the abnormal causal path, thereby guiding subsequent control strategies.

[0086] In some embodiments of the present invention, the time difference obtained by subtracting the time point of the abnormal occurrence of the cap from the time point of the occurrence of the regulatory behavior of the second potential intermediate node is multiplied by a preset attenuation coefficient, and then an exponential operation with the natural constant as the base is performed to obtain the time matching degree score; if the time difference between the time point of the abnormal occurrence of the cap and the time point of the occurrence of the regulatory behavior of the second potential intermediate node is greater than 72 hours, the corresponding potential association path is abandoned.

[0087] Specifically, the time difference is the time interval between the occurrence of cap anomalies and the occurrence of the regulatory behavior of the second potential intermediate node, typically calculated in hours. The preset decay coefficient is a pre-defined constant used to control the rate at which the time matching degree decays as the time difference increases. It is an exponential calculation with a base of a natural constant, for example, it can be used... The form is , where e is the natural constant, k is the preset attenuation coefficient, and Δt is the time difference.

[0088] The exponential decay model effectively reflects the timeliness and effectiveness of regulatory behaviors on cap anomalies. Specifically, the more recent the regulatory behavior, the stronger its correlation; conversely, the correlation decreases exponentially with the time elapsed for more distant regulatory behaviors. Furthermore, if the time difference between the occurrence of the cap anomaly and the occurrence of the regulatory behavior at the second potential intermediate node is greater than 72 hours, the regulatory behavior is considered to have extremely low or no correlation with the current cap anomaly. Therefore, the corresponding potential correlation path is directly discarded and not included in the correlation strength score calculation, to avoid interference from irrelevant information in determining the causal path of the anomaly.

[0089] This scheme introduces an exponential decay model to calculate the time-matching score, enabling a more scientific and accurate quantification of the temporal correlation strength between regulatory behaviors and cap abnormalities. When the time of occurrence of the regulatory behavior and the time of occurrence of the cap abnormality are close, the time difference is small, the exponential calculation result is large, thus assigning a high time-matching score, indicating a strong correlation. Conversely, when the time difference is large, the exponential calculation result decreases rapidly, causing the time-matching score to decrease, reflecting a weakened correlation. This non-linear decay mechanism, compared to simple linear or threshold judgments, more accurately reflects the dynamic changes in the influence of environmental factors during biological growth. Furthermore, setting a 72-hour time threshold effectively filters out regulatory behaviors that are too far in time and almost impossible to have a causal relationship with the current cap abnormality, thereby reducing computational load, improving system efficiency, and avoiding misclassification of irrelevant historical data as the cause of the abnormality.

[0090] Suppose that an abnormal cap color is detected in a mushroom bed at time T0. During the reverse lookup process, the system identifies three potential regulatory behaviors: 1. Control behavior A: Humidity was adjusted 12 hours before T0.

[0091] 2. Regulation behavior B: Temperature adjustment was carried out 48 hours before T0.

[0092] 3. Regulation behavior C: Light intensity was adjusted 96 hours before T0.

[0093] Set the preset attenuation coefficient k to 0.05.

[0094] For regulatory behavior A, the time difference Δt = 12 hours. The time matching score = e^(-0.05 * 12) = e^(-0.6) ≈ 0.5488.

[0095] For regulatory behavior B, the time difference Δt = 48 hours. The time matching score = e^(-0.05 * 48) = e^(-2.4) ≈ 0.0907.

[0096] For regulatory behavior C, the time difference Δt = 96 hours. Since 96 hours is greater than the preset threshold of 72 hours, potential correlation paths related to regulatory behavior C are directly abandoned, and their time matching score is considered to be 0.

[0097] In this way, regulatory behavior A, due to its occurrence time being closer to the time of the cap abnormality, obtained a higher time-matching score, indicating a stronger association with the abnormality. The association of regulatory behavior B was significantly reduced. Regulatory behavior C, however, was excluded because its occurrence time was too long, avoiding erroneous attribution to irrelevant historical events. Therefore, the system can more effectively focus on recent and potentially relevant regulatory behaviors, thereby improving the accuracy of identifying abnormal causal paths.

[0098] In some embodiments described above in this application, an abnormal control strategy based on an abnormal causal path is proposed for generation and adjustment. However, in its implementation, if the specific information on the abnormal causal path is not fully utilized, the generated control strategy may be inaccurate or inefficient, thus affecting the effectiveness of edible fungi growth management. Therefore, this application further proposes the following steps for generating and adjusting an abnormal control strategy based on an abnormal causal path: Based on each node in the abnormal causal path, identify the node that causes the abnormal cap color; An abnormal control strategy is generated based on the node content that causes the abnormal cap color on the abnormal causal path. The abnormal control strategy is sent to the control equipment in the corresponding area, and the control equipment is controlled to perform control according to the abnormal control strategy.

[0099] Specifically, after obtaining the abnormal causal path of the target edible fungus, an in-depth analysis of this path is required. Identifying the node causing the abnormal cap color based on each node in the abnormal causal path involves analyzing the direct or indirect correlation between each node in the path (e.g., variety, growth environment parameters, regulatory behavior, type and location of regulatory equipment, biological characteristics, etc.) and the abnormal cap color to identify the most direct or fundamental pathogenic factor. For example, if the path shows that "excessively high temperature" leads to "excessively rapid moisture evaporation," which in turn leads to "abnormal cap color," then either "excessively high temperature" or "excessively rapid moisture evaporation" could be identified as nodes causing the abnormal cap color.

[0100] Furthermore, based on the content of the nodes on the abnormal causal path that cause the abnormal cap color, anomaly control strategies are generated. This means that once a pathogenic node is identified, the system will automatically or semi-automatically generate corresponding control instructions based on the specific content of that node (e.g., excessively high temperature, excessively low humidity, insufficient light, etc.). For example, if the pathogenic node is determined to be excessively high temperature, the generated control strategy might include reducing the ambient temperature by X degrees Celsius; if it is determined to be excessively low humidity, the strategy might be to increase the ambient humidity by Y%. These strategies aim to directly correct the root cause of the abnormality.

[0101] Furthermore, sending the abnormal control strategy to the corresponding control equipment in the area, and controlling the control equipment to perform control according to the abnormal control strategy, involves transmitting the generated specific control instructions to the corresponding control equipment in the area where the target edible fungus is located via a network or other communication methods. For example, if the strategy is to lower the temperature, the instruction will be sent to the air conditioning system or ventilation equipment in that area; if the strategy is to increase humidity, the instruction will be sent to the humidifier. After receiving the instruction, the control equipment will automatically execute the corresponding operation to achieve precise adjustment of environmental parameters.

[0102] In some preferred embodiments, suppose that in a certain edible mushroom cultivation area, the caps of target edible mushrooms are identified as having abnormal color, and the causal path of the abnormality is determined through an association path reasoning model as: excessively high ambient temperature -> excessively rapid evaporation of fungal moisture -> abnormal cap color. Based on this causal path, the system first identifies "excessively high ambient temperature" as the key node causing the abnormal cap color. Subsequently, based on the "excessively high ambient temperature" node, a specific abnormality control strategy is generated, such as "reducing the ambient temperature of this area by 2 degrees Celsius." Finally, this control strategy is sent to the temperature and humidity control equipment in the area, such as a smart air conditioner or ventilation system, to control it to adjust the temperature according to the strategy, thereby effectively solving the problem of abnormal cap color.

[0103] Furthermore, after adjusting according to the above-mentioned abnormal control strategy, the process also includes monitoring the growth of the target edible fungi, including: Obtain a panoramic image of the target edible fungus's mushroom bed, input it into the target edible fungus cap anomaly recognition model, and determine whether there are any abnormal cap color areas in the panoramic image of the target edible fungus's mushroom bed based on the output results; If, after the third preset time following the implementation of the control strategy, the abnormal color area of ​​the cap in the panoramic image of the target edible fungus disappears, then the monitoring of the growth of the target edible fungus will be stopped. If the abnormal color area of ​​the cap in the panoramic image of the target edible fungus bed does not disappear after the third preset time following the implementation of the control strategy, an alarm signal will be issued.

[0104] Specifically, the steps for monitoring the growth of target edible fungi aim to continuously track the status of the fungi after the implementation of the control strategy. Acquiring panoramic images of the fungal bed involves periodically or as needed taking images of the entire fungal bed using image acquisition equipment, such as a high-definition camera, to comprehensively capture its growth status. The acquired panoramic images are then input into the aforementioned target edible fungi cap anomaly recognition model. This model is a pre-trained intelligent model capable of accurately identifying areas of abnormal cap color. The output of this model determines whether abnormal cap color areas exist in the panoramic image, thereby assessing whether the control strategy has produced the expected results.

[0105] After the third preset time following the implementation of the control strategy, the abnormality of the mushroom caps will be assessed again. This third preset time can be understood as the reasonable time period required for the control strategy to take effect and show its effects; for example, it can be set to 24 hours, 48 ​​hours, or 72 hours, to allow the control strategy sufficient reaction time. If, after this time point, the abnormal color area of ​​the mushroom caps disappears in the panoramic image of the target edible fungus bed, it indicates that the control strategy has successfully resolved the abnormality problem, and monitoring of the target edible fungus's growth can be stopped to save system resources. Conversely, if the abnormal color area of ​​the mushroom caps has not disappeared, it means that the current control strategy may be ineffective or poorly performing. In this case, the system will issue an alarm signal to remind management personnel or the system to take further intervention measures, such as reassessing the causal path of the abnormality, adjusting the control strategy, or conducting manual inspections.

[0106] This application's solution effectively addresses the problem of untimely verification of control effects in the basic solution by introducing a post-control monitoring mechanism. Specifically, after the abnormal control strategy is executed, the system does not simply terminate the process. Instead, it continuously acquires panoramic images of the mushroom bed and analyzes them using a target edible mushroom cap anomaly recognition model to assess in real time whether the abnormal cap area has disappeared. This proactive feedback mechanism ensures that the effectiveness of the control strategy can be quantified and verified. The inclusion of a third preset timeframe as a window for effect evaluation allows the system sufficient time for the control strategy to take effect, avoiding premature or delayed judgments. When the abnormal cap area disappears, the monitoring stops, avoiding unnecessary resource consumption; while when the abnormal area persists, the alarm signal triggers subsequent intervention processes in a timely manner, preventing further deterioration of the problem, thus forming a closed-loop, adaptive edible mushroom growth management system.

[0107] Furthermore, after adjusting according to the above-mentioned anomaly control strategy, the method further includes adding the abnormal causal path of the target edible fungus to the association path reasoning model in the form of nodes and edges. Specifically, the path formed by connecting the starting node and the first potential intermediate node in the abnormal causal path of the target edible fungus is added to the association path reasoning model as a first-level association path; the path formed by connecting the first potential intermediate node and the second potential intermediate node in the abnormal causal path of the target edible fungus is added to the association path reasoning model as a second-level association path.

[0108] Specifically, the abnormal causal path is determined by calculating the association strength scores of potential related paths and selecting the path with the highest association strength score, thus identifying the root cause chain leading to the abnormality of the target edible mushroom cap. This abnormal causal path consists of a series of interconnected nodes and edges, tracing from the abnormal cap phenomenon to growth environment parameters, regulatory behaviors, and even the location of regulatory equipment. Adding this abnormal causal path as nodes and edges to the association path reasoning model means structurally integrating the actual, verified causal relationships into the knowledge graph, thereby enriching and improving the content of the knowledge graph. Here, the starting node typically refers to the abnormal phenomenon reflected in the cap abnormality identification result, the first potential intermediate node refers to the growth environment parameters directly associated with the starting node, and the second potential intermediate node refers to the regulatory behaviors that cause or affect the first potential intermediate node. The path connecting the starting node and the first potential intermediate node is defined as a first-level association path, aiming to strengthen the understanding of the association between the abnormal phenomenon and direct environmental factors; the path connecting the first potential intermediate node and the second potential intermediate node is defined as a second-level association path, aiming to clarify the causal relationship between environmental factors and specific regulatory behaviors. This hierarchical approach to adding information allows for more precise updates to the relationships within the knowledge graph.

[0109] This application's solution achieves continuous learning and self-improvement of the knowledge graph by dynamically feeding back and adding identified and verified abnormal causal paths to the association path reasoning model. When the system successfully identifies the causal path of the cap anomaly and takes control measures, the verified causal chain is structurally stored in the knowledge graph. Thus, the knowledge graph is no longer static but evolves and enriches with each anomaly handling process. This mechanism enables the system to reason based on a more comprehensive and accurate knowledge graph when facing similar or related anomalies in the future, thereby improving the efficiency and accuracy of reasoning. The distinction between primary and secondary association paths further refines the causal chains in the knowledge graph, allowing the system to more accurately understand the relationships between different levels of technical features.

[0110] like Figure 2As shown, this application also discloses a knowledge graph-based edible fungus growth management system, including: The anomaly identification module 210 is used to identify the abnormalities of the caps of target edible fungi in a multi-regional mixed cultivation environment and generate the corresponding cap anomaly identification results. The potential association path determination module 220 is used to construct an association path reasoning model corresponding to the target edible fungus based on a knowledge graph, and obtain the potential association path corresponding to the cap anomaly identification result based on the cap anomaly identification result and the association path reasoning model. The abnormal causal path determination module 230 is used to calculate the association strength score of each potential association path and select the potential association path with the highest association strength score as the abnormal causal path of the target edible fungus. The control module 240 is used to generate an abnormal control strategy based on the abnormal causal path and to adjust according to the abnormal control strategy.

[0111] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A knowledge graph-based method for the management of edible fungi growth, characterized in that, include: In a multi-regional mixed cultivation environment for edible fungi, abnormalities in the caps of target edible fungi are identified, and corresponding cap abnormality identification results are generated. Based on the knowledge graph, an association path reasoning model corresponding to the target edible fungus is constructed. Based on the cap anomaly identification result and the association path reasoning model, the potential association path corresponding to the cap anomaly identification result is obtained. Calculate the association strength score of each potential association path, and select the potential association path with the highest association strength score as the abnormal causal path of the target edible fungus. Based on the abnormal causal path, an abnormal control strategy is generated and adjusted according to the abnormal control strategy.

2. The method for managing the growth of edible fungi based on knowledge graphs according to claim 1, characterized in that, In a multi-regional mixed cultivation environment for edible fungi, abnormalities in the caps of target edible fungi are identified, and corresponding cap abnormality identification results are generated, including: Obtain a panoramic image of the target mushroom bed for the target edible fungus; The panoramic image of the target mushroom bed is input into a pre-trained target edible mushroom cap anomaly recognition model. The output result is used to determine whether there are abnormal cap color areas in the panoramic image of the target mushroom bed. The pre-trained target edible mushroom cap anomaly recognition model is obtained by training the target edible mushroom cap anomaly recognition model on images of healthy target edible mushroom caps and images of target edible mushroom caps with identified diseased and abnormal caps. For a panoramic image of a target mushroom bed with abnormal cap color, determine the number of abnormal cap color areas, the color deviation value of each abnormal cap color area, the proportion of abnormal areas, the duration, and the sub-image of each abnormal cap color area, and generate the cap anomaly recognition result.

3. The method for managing the growth of edible fungi based on knowledge graphs according to claim 1, characterized in that, The association path reasoning model for the target edible fungus is a graph database query engine that includes the knowledge system of the target edible fungus; wherein, The nodes in the target edible fungus knowledge system are the variety of the target edible fungus, growth environment parameters, regulatory behavior, type and location of regulatory equipment, and biological characteristics of the target edible fungus. In the target edible fungi knowledge system, the edges represent the relationships between different nodes, and the edges formed by connecting the same node to other different nodes have different weights.

4. The method for managing the growth of edible fungi based on knowledge graphs according to claim 2, characterized in that, Based on the cap anomaly identification result and the association path reasoning model, the potential association path corresponding to the cap anomaly identification result is obtained, including: In the aforementioned association path reasoning model, the node related to the cap anomaly identification result is selected as the starting node; A reverse lookup is performed in the knowledge system to find growth environment parameters that are associated with the starting node, which are then used as the first potential intermediate nodes. For each first potential intermediate node obtained by reverse query, the reverse query in the knowledge system causes or affects the regulatory behavior of the corresponding first potential intermediate node, which is then used as a second potential intermediate node. Based on each of the second potential intermediate nodes obtained from the query, a reverse query is performed in the knowledge system to determine the control device and its location for executing the second potential intermediate node, which serves as the termination node. The potential associated path includes the termination node, the second potential intermediate node, the first potential intermediate node, and the starting node.

5. The method for managing the growth of edible fungi based on knowledge graphs according to claim 4, characterized in that, In the step of calculating the association strength score for each of the potential association paths, the calculation logic includes: If the number of nodes in a potential association path is less than a preset first number of nodes, then the first association strength score for the corresponding potential association path is assigned as the first preset association strength score; if the number of nodes in a potential association path is greater than or equal to the preset first number of nodes but less than a preset second number of nodes, then the first association strength score for the corresponding potential association path is assigned as the second preset association strength score; if the number of nodes in a potential association path is greater than the preset second number of nodes, then the first association strength score for the corresponding potential association path is assigned as the third preset association strength score; wherein, the first preset association strength score is greater than the second preset association strength score, which is greater than the third preset association strength score. In the knowledge system, a preset node confidence level is set for each node, a preset edge confidence level is set for each edge, and the second association strength score of the potential associated path is the product of the preset node confidence level of all nodes and the preset edge confidence level of each edge. Compare the occurrence time of cap abnormality with the occurrence time of the regulatory behavior of the second potential intermediate node. If the occurrence time of the regulatory behavior of the second potential intermediate node is within a first preset time range before the occurrence time of cap abnormality, the third association strength score of the potential association path is given as the time matching score calculated based on the occurrence time of cap abnormality and the occurrence time of the regulatory behavior of the second potential intermediate node. Within a second preset time range before and after the abnormality of the cap of the target edible fungus, whether the control device in the area outside the area where the target edible fungus is located exhibits control behavior. If it does, the fourth association strength score of the potential association path is given as the control association score. If it does not, the fourth association strength score of the potential association path is given as 0. The calculated association strength score for any potential association path is equal to the first association strength score plus the second association strength score plus the third association strength score plus the fourth association strength score.

6. The method for managing the growth of edible fungi based on knowledge graphs according to claim 5, characterized in that, The time difference obtained by subtracting the time point of the abnormal occurrence of the cap from the time point of the occurrence of the regulatory behavior of the second potential intermediate node is multiplied by a preset attenuation coefficient, and then an exponential operation with the natural constant as the base is performed to obtain the time matching degree score. If the time difference between the occurrence of cap abnormality and the occurrence of the regulatory behavior of the second potential intermediate node is greater than 72 hours, the corresponding potential association path is abandoned.

7. The method for managing the growth of edible fungi based on knowledge graphs according to claim 1, characterized in that, Based on the aforementioned abnormal causal path, an anomaly control strategy is generated, and adjustments are made according to the aforementioned anomaly control strategy, including: Based on each node in the abnormal causal path, identify the node that causes the abnormal cap color; An abnormal control strategy is generated based on the node content that causes the abnormal cap color on the abnormal causal path. The abnormal control strategy is sent to the control equipment in the corresponding area, and the control equipment is controlled to perform control according to the abnormal control strategy.

8. The method for managing the growth of edible fungi based on knowledge graphs according to claim 7, characterized in that, Following the adjustment steps according to the aforementioned abnormal control strategy, the method further includes a step of monitoring the growth of the target edible fungi, including: A panoramic image of the target edible fungus's mushroom bed is obtained, and the panoramic image of the target edible fungus's mushroom bed is input into a pre-trained target edible fungus cap anomaly recognition model. The output result is used to determine whether there are abnormal cap color areas in the panoramic image of the target edible fungus's mushroom bed. The pre-trained target edible fungus cap anomaly recognition model is obtained by training the target edible fungus cap anomaly recognition model on images of healthy caps of the target edible fungus and images of caps of the target edible fungus that have been identified as diseased and abnormal. If, after the third preset time following the execution of the control strategy, the abnormal color area of ​​the cap in the panoramic image of the target edible fungus disappears, then the monitoring of the growth of the target edible fungus will cease. If the abnormal color area of ​​the cap in the panoramic image of the target edible fungus bed does not disappear after the third preset time following the implementation of the control strategy, an alarm signal will be issued.

9. The knowledge graph-based edible fungus growth management method according to claim 1, characterized in that, After the step of adjusting according to the aforementioned anomaly control strategy, the following is also included: The abnormal causal path of the target edible fungus is added to the association path reasoning model in the form of nodes and edges; wherein, The path formed by connecting the starting node and the first potential intermediate node in the abnormal causal path of the target edible fungus is added to the association path inference model as a first-level association path. The path formed by connecting the first potential intermediate node and the second potential intermediate node in the abnormal causal path of the target edible fungus is added to the association path reasoning model as a secondary association path.

10. A knowledge graph-based edible fungus growth management system, characterized in that, include: The anomaly identification module is used to identify abnormalities in the caps of target edible fungi in a multi-regional mixed cultivation environment and generate corresponding cap anomaly identification results. The potential association path determination module is used to construct an association path reasoning model corresponding to the target edible fungus based on a knowledge graph, and to obtain the potential association path corresponding to the cap anomaly identification result based on the cap anomaly identification result and the association path reasoning model. An abnormal causal path determination module is used to calculate the association strength score of each potential association path and select the potential association path with the highest association strength score as the abnormal causal path of the target edible fungus. The control module is used to generate an anomaly control strategy based on the abnormal causal path and to adjust according to the anomaly control strategy.

Images